web
You’re offline. This is a read only version of the page.
close
Skip to main content
Community site session details

Community site session details

Session Id :
Copilot Studio - Copilot Studio Skills Development
Suggested answer

Generative Orchestration shows unmasked draft before my PII topic replaces it — how to prevent?

(1) ShareShare
ReportReport
Posted on by 2

Hi all 👋

I’m on Copilot Studio with Generative Orchestration ON. I use a custom topic triggered by “AI response generated” to PII-mask the model’s draft and replace it.

 

Issue

After “Fetching knowledge,” the unmasked draft briefly appears, then my masked message replaces it. I need to prevent any unmasked text from showing at all.

 

What I’ve tried:

  • In the custom topic:
    Run a prompt (PII) with input = System.Response.FormattedText → returns sanitizedContent.
    Send only sanitizedContent.
    System.ContinueResponse = false (tested as first and last step).
    With and without Power Automate (when used, returns values only, no user message).
    Also tried Search knowledge sources inside the topic (mask then send).


  1. How can I stop the generative draft from streaming so only my masked message is shown?


  2. Is there a pattern to guarantee zero exposure of unmasked content in Generative mode?
     

 

 

 

Thanks a lot in advance! 🙏

 

 
 
Categories:
I have the same question (0)
  • Suggested answer
    Romain The Low-Code Bearded Bear Profile Picture
    1,155 Super User 2025 Season 2 on at
    Generative Orchestration shows unmasked draft before my PII topic replaces it — how to prevent?
    hello
     
    this happen to me when the conversationnal boosting "think" the answer is not enough or out of bound.
    You could disable it by going to Topics->System->conversationnal boosting and disable. Then you will have few error inside the agent because few other system topics use it, correct / remove the part and then go.
     
    This system topic "conversationnal boosting" is a relic of the past when GenAI was not default mode (the classic one was). it was very very usefull at this time.
    With gen AI orchestration, knolwedge are triggered automatically after topics, so conversationnal boosting is not needed anymore in real life agent, it's just a "safe net" to help.
     
    most of the time, it solve my problem.
     
    If not : check the general knowledge if it's disable or enable (could influence) and in gen AI settings, you could check default AI model and change (it could influence too)
     
    hope this help :) if yes please check the answer button to make the post green, it's important for the community and search engine :)

Under review

Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.

Helpful resources

Quick Links

Responsible AI policies

As AI tools become more common, we’re introducing a Responsible AI Use…

Tom Macfarlan – Community Spotlight

We are honored to recognize Tom Macfarlan as our Community Spotlight for October…

Leaderboard > Copilot Studio

#1
Romain The Low-Code Bearded Bear Profile Picture

Romain The Low-Code... 523 Super User 2025 Season 2

#2
Michael E. Gernaey Profile Picture

Michael E. Gernaey 378 Super User 2025 Season 2

#3
DAnny3211 Profile Picture

DAnny3211 132

Last 30 days Overall leaderboard